Capability

Singing Generation

Generate a vocal performance from a prompt and lyrics, optionally guided by a reference voice sample or a reusable saved voice clone. Available in v5, v6, and v7.

Compat route available

POST /v{5,6,7}/generate/singing covers generate_singing with old-style field names. This is a new URL, not a literal recreation — the old singing endpoint was version-less (/v2/generate/ai-singing-gen), predating the v5/v6/v7 family entirely, so this path is offered for consistency with the rest of the compat surface rather than resurrecting the old one. See Compat Routes.

Quickstart

POST/v7/generate/singingcompat route, simplest way to start

curl -X POST https://apiv2.soundverse.ai/v7/generate/singing \ -H "Authorization: Bearer sksoundverse_..." \ -H "Content-Type: application/json" \ -H "Idempotency-Key: singing-req-001" \ -d '{ "styleOfMusic": "warm soul, female alto", "lyrics": "walking through the city lights...", "vocal_id": "sv_voice_..." }'

The compat route maps styleOfMusicprompt and audioUrlvocal_reference; it also accepts a saved vocal_id from Voice Cloning. The saved voice must belong to your account and have active consent; do not send a provider voice ID. Response: 201 {"job_id": "...", "status": "queued"}.

generate_singing — native payload

POST/v1/generationstool_id for (model, "generate_singing")

FieldTypeRequiredDefaultDescription
promptstringNo""Vocal style — mood, timbre, genre. Up to 1024 characters.
lyricsstringNo""Lyrics to sing. Up to 3000 characters.
vocal_referencefile inputNoAn optional voice sample to sing in a similar voice. mp3/m4a, ~30s, up to 15MB.
vocal_idstringNo""Optional saved Soundverse voice ID (sv_voice_...) created by Voice Cloning. Must be owned by your account and have active consent.
versionsintegerNo1Number of vocal versions to generate. 1–3.

The underlying worker fleet also has a sibling similar_singing operation, but it is not currently enterprise-priced or seeded — it isn’t available through this API today.

Results and playback

Poll GET /v1/generations/{job_id} until it is completed. When you supplied a saved voice, output.metadata_json.final_payload.vocal_id echoes the same sv_voice_... ID, confirming that it was used for the generation.

Listen from Soundverse Library

The completed audio is saved to the API key owner’s Soundverse Library. Open it while signed in as that account to play or download the result. Do not openoutput.assets[].url directly: it is a private storage locator rather than a public playback URL.

For backend download, call GET /v1/files/{file_id}/download. Enterprise API outputs are hidden from public/profile surfaces by default.

Pricing

VersionRoyalty-FreeStandardDistributionSyncMaster
v5 / v6 / v7$0.09$0.14$0.18$0.36$0.90

USD per generation, by license tier.